Due to unit outage and fluctuation of renewable power generation and load, scheduling reasonable amount of spinning reserve is important for maintaining safe and economical operation of microgrids. Based on day-ahead forecast errors models of wind power, photovoltaic power and load, the uncertainty distribution models of wind power and photovoltaic power are built by total probability formula that comprehensively considers forecast errors and unit outage. After discretizing wind power, photovoltaic power and load uncertainty distributions, the discrete uncertainty distribution model of microgrid power is generated combining with the outage probability distributions of dispatchable units. Then taking minimized microgrid total social cost as objective, a computing model of optimal spinning reserve in microgrids is presented considering uncertain factors of microgrid power and supplying spinning reserve to the main grid. Finally an example confirms the feasibility of the proposed model after optimizing the spinning reserve by mixed-integer genetic algorithm.